DeepFake Detection using Transfer Learning

A deepfake is a computer-generated fake image or video that combines images to create a new image or video that depicts an event, comment, or activity that did not actually occur. This has become a real problem nowadays to decide the originality of a video. For the same reason we are trying to create a machine learning model using transfer learning which will help us to distinguish between different videos to decide which video is real and which one is fake. For that we are using four different models and comparing their results. The overall best model is InceptionResNetV2 considering its training time and accuracy. In our results the InceptionResNetV2 performs best and gives an accuracy of 97.1 percent.